Skip to main content

Module intervals

Module intervals 

Source
Expand description

The readings’ 90% intervals held to the synthetic flights’ truth (ADR-257 §1–2, ADR-308).

Each flight’s own log is read by hpr_flightdata::readings, and each reading that states a 90% interval (hpr_flightdata::readings::interval) is compared with the flight’s truth: below the interval, inside it, or above it. The coverage test of ADR-257 §1, calibrated intervals, takes the three counts over a set, each with its two-sided 99.9% Clopper–Pearson band: the two tails’ bands must each hold 5%, and the band of the count inside must hold 90%. At 1,000 readings a tail passes with 29 to 74 misses. Three probes must each fail it: every interval narrowed to 0.8 of its width about its middle, widened to 1.25, and shifted down by one stated standard error, the width over 2 × 1.645. For normal errors those hold the truth about 81% and 96% of the time, and the shifted one misses above about 26% of the time.

A floor’s truth may lie any amount above it, so it states only a lower end, tested on its lower tail alone: the band on the share of truths below it must hold 5%, raised one stated standard error and lowered two it must not (FloorCheck). A top acceleration’s floors come from logs clipping at 8 g, its top scaled to anywhere the reader can call it clipped and the part may not have clipped (top_over_clip), and its top as flown, which may miss less but not more.

The intervals’ shares are fitted on Set::Calibration (fitted); the test runs on Set::A and Set::B, which the shares never see.

Structs§

Band
A share’s two-sided 99.9% Clopper–Pearson band.
Check
One reading’s check on one set.
Compared
One reading’s interval against its truth.
ComparedFloor
One floor’s lower end against its truth.
Counts
Readings whose truth lay below their interval, inside it and above it.
FlightSaid
A flight’s index, its design, logger and day’s air, and what its own log said of each of STATED.
FloorCheck
One reading’s floors on one set.
FloorTest
The test of a floor’s lower end, the coverage test’s lower tail: the band on the share of truths below it must hold 5%.
Test
The coverage test on one set of counts: each tail’s band must hold 5%, the inside’s 90%.

Enums§

AirBand
A band of the day’s air, K from the 1976 standard atmosphere. A barometric height is off by about ΔT/T of itself, so the height’s errors, and its misses, gather at the ends of the synthetic flights’ ±15 K: a warm day reads low, a cold one high.
FloorProbe
A change to every floor’s lower end that its test must catch.
Probe
A change to every interval that the coverage test must catch.
Said
What one log said of one reading.

Constants§

FLOORED
The readings that state a lower end where they are a floor, in the report’s order.
INSIDE_SHARE
The share of readings whose truth the interval must hold.
LEAST_FLOORS
The fewest floors whose test a probe can fail: of 148, no truth below still holds 5%.
STATED
The readings that state a 90% interval, in the report’s order.

Functions§

check
quantity’s check over flights.
fitted
quantity’s shares fitted again on flights, which should be Set::Calibration’s; None when fewer than 20 state an interval or a share isn’t a number.
fitted_floor
quantity’s floor shares fitted again on flights; None when fewer than 20 floors are read or a share isn’t a number.
floor_shares
The library’s floor shares for quantity; None for a reading that is never a floor here.
said
What flight’s own log says of each of STATED, against its truth.
shares
The library’s shares for quantity; None for a reading that states no interval yet.
three_figures
x to three significant figures, as the library’s shares are written.
top_over_clip
Flight index’s top over the clip on the log its lower end is fitted on, for a part whose scale error runs to ±scale. It spreads evenly over every top the reader can call clipped and the part may not have clipped. At the bottom, the reader calls a top clipped from GRAVITY_TOLERANCE_G under the range, and the part’s gain and ACCEL_BIAS_G can carry a truth up to that: (1 − (tolerance + bias)/range)/(1 + scale). At the top, past 1/(1 − scale) every such part clips and the truth lies above the reading. A lower top simply reads whole, and doesn’t count. Each flight’s place is its own, the golden ratio’s multiples from a start each set’s seed sets, so no design’s thrust sets it and no two sets share it.
withheld_by_rule
Whether quantity withheld for reason from a log with a barometer is the reading’s own rule and not a reading lost: the top speed is withheld at or above hpr_flightdata::readings::TRANSONIC_SPEED_M_S, 227.7 m/s, where the barometer’s port may read the transonic flow over the nose. The rule looks at the reading, not the truth, so the test holds the readings it lets through, as a user sees them. A burnout taken from that speed goes with it, and a log with no accelerometer gives no top acceleration.